1. What is rstanarm and why would you use it?

• it’s an R interface to stan

• it’s better than rstan, because (according to rstanarm webpage)

  • models are specified with formula syntax,

  • data is provided as a data frame, and

  • additional arguments are available to specify priors.

• in a nutshell, rstanarm let’s you estimate various Bayesian models and examine them with shinystan.

2. A “Hello World” example

Let’s begin our adventure with rstanarm with a package on which rstanarm is built, i.e. rstan. Why was rstanarm even created? It provides a nice R interface for many rstan fucntions, in particular, you do not have to keep your model definition in a separate “.stan” file nor in a string, but you write it just as you did in R, with a function + formula syntax.

In the following examples we will calculate coefficients of the simplest statistical model, which is linear regression.

rstan example

Let’s see how we can estimate coefficients of linear regression in rstan.

based on http://thestatsgeek.com/2015/06/08/my-first-foray-with-stan/

model_code <- "
data {
  int N;
  vector[N] x;
  vector[N] y;
}
parameters {
  real alpha;
  real beta;
  real sigma;
}
model {
  y ~ normal(alpha + beta * x, sigma);
}
"

set.seed(123)
n <- 100
x <- rnorm(n)
y <- x + rnorm(n)
mydata <- list(N = n, y = y, x=x)

library(rstan)
fit <- stan(model_code = model_code, data = mydata, 
            iter = 1000, chains = 4)
fit

summary(lm(y~x))

Having the coefs estimated, let’s see the results in a graphical, interactive form:

base on http://mc-stan.org/shinystan/articles/shinystan-package.html

library(shinystan)
my_sso <- launch_shinystan(fit)

What happened here?

• we defined model_code, in which we provided all the important information about estimation of the model; we could have save it in a .stan file;

• we created a sample dataset and saved it into a list mydata;

• we fitted the model with stan;

• we may compare it’s results with a regular model estimated with lm();

• we launched the shinystan GUI to look into the model.

Concluding, modelling in rstan does not really resemble R programming language that we use every day.

rstanarm example

Let’s use the dataset from the previous section (rstan example) for estimating coefficients of linear regression.

based on http://m-clark.github.io/workshops/bayesian/04_R.html#rstanarm

library(rstanarm)

mydata_df <- data.frame(y = y, x = x)
fit2 <- stan_glm(y ~ x, data = mydata_df, iter=1000, chains = 4)
fit2

library(shinystan)
launch_shinystan(fit2)

What happened here?

• we loaded rstanarm package;

• we created a data.frame, which contains our vectors from previous section;

• we fitted a generalised linear model with glm()

• and launched shinystan to view it in GUI.

Rstanarm provides a much more familiar syntax, comparing to rstan.

3. Useful links

rstan

• example of least squares in rstan

• RStan

rstanarm

• a very good introductory article

• tutorial/documentation/vignette

• other vignettes

• rstanarm + mixed effects

• datacamp course of rstanarm

• User-friendly Bayesian regression modeling: A tutorial with rstanarm and shinystan

Bayes

• Doing Bayesian Data Analysis - probably the best book to learn the Bayesian philosophy